Rainer Jung wrote:
On 07.03.2009 15:23, Mladen Turk wrote:
Rainer Jung wrote:
On 07.03.2009 09:00, mt...@apache.org wrote:
After 10 seconds or after 60 seconds: I think 60 seconds is pretty
long, but I would accept as a compromise :)
Let's use the recover_wait_time here, or if you still think
this is huge value I'll handle the recover_wait_time / 2 :)
OK, will implement on Monday. I'll take recover_wait_time / 2 as a
default and might make configurable (after thinking about how much
trouble an admin could produce when using a bad configuration).
Then we are almost through, did you have a look at the force recovery
patch? I think that one is safe.
Isn't that related to the first one?
However I still think this is edge case where OS socket
implementation doesn't detect the cable is pulled out.
Using cping and watchdog is the proper solution to handle this
without making normal operations broken, and this patch will did that.
And finally: the global state setting for the new APR_PROTOCOL error.
Why would you want to continue talking to a node that send back garbage?
Because garbage or protocol error doesn't mean the node is down.
The JVM might just be in full GC and socket times out.
Each connection should decide by itself.
Regards
--
^(TM)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@tomcat.apache.org
For additional commands, e-mail: dev-h...@tomcat.apache.org
